Exampel-usage of XCSVM

This script shows how one can use xcsvm. It provides shell commands and the according output. When running the example one should get similar outputs.

The following cases are shown:

Training and saving a model and then predicting.
Training and predicting in one shot.
Training with two cores.
Training with two processes.

As a data set we use LSHTC-small. All the examples can also be used from the file examples.bash.

Paths

We assume that you launch ipython notebook in the example folder, i.e. the following relative paths are correct.



In [1]:

    
TRAIN_DATA = "../xcsvm/xcsvm/tests/data/lshtc1_dryrun_task1_tfidf_l2_norm_applied.train.libsvm"
TEST_DATA = "../xcsvm/xcsvm/tests/data/lshtc1_dryrun_task1_tfidf_l2_norm_applied.test.libsvm"

# The following python interpreter should have xcsvm installed.
# If not used the run.py script will try to run the source code directly.
# In this case all the needed packages need to be installed.
PYTHON = "/usr/bin/python2.7"

Training and saving model

We train and store the model.

The run.py script will always store the model when '--model_dir' is provided.



In [2]:

    
! rm -rf ./model
! $PYTHON ../scripts/run.py ww_sparse --train_data $TRAIN_DATA --model_dir ./model --options folds=5 group_count=64 -v 1









    



[Rank 0] 2017-03-30 15:05:06,277 -    base.py -    INFO: Solver uses testing mode.
[Rank 0] 2017-03-30 15:05:06,334 -     run.py -    INFO: Loading train data.
[Rank 0] 2017-03-30 15:05:07,124 -     run.py -    INFO: Start training.
[Rank 0] 2017-03-30 15:05:07,124 -    base.py -    INFO: Use C=1 and epsilon=0.1
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
[Rank 0] 2017-03-30 15:05:07,124 -    base.py -    INFO: Using 1 MPI processes and 1 MP threads.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.ww <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.ww' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/ww.so'>
[Rank 0] 2017-03-30 15:05:07,152 -    base.py -    INFO: Remove 0 samples where K is 0.
[Rank 0] 2017-03-30 15:05:07,183 -      ww.py -    INFO: Use 64 groups.
[Rank 0] 2017-03-30 15:05:07,218 -    base.py -    INFO: Mem allocations:
[Rank 0] 2017-03-30 15:05:07,219 -    base.py -    INFO:   Total: 522.870 MB
[Rank 0] 2017-03-30 15:05:10,192 -    base.py -    INFO: Iter:          1, Max. Viol.:   2.273799, Time: 0:00:02.973290
[Rank 0] 2017-03-30 15:05:13,033 -    base.py -    INFO: Iter:          2, Max. Viol.:   2.343154, Time: 0:00:05.814228
[Rank 0] 2017-03-30 15:05:15,865 -    base.py -    INFO: Iter:          3, Max. Viol.:   1.059850, Time: 0:00:08.646201
[Rank 0] 2017-03-30 15:05:18,679 -    base.py -    INFO: Iter:          4, Max. Viol.:   0.730385, Time: 0:00:11.459940
[Rank 0] 2017-03-30 15:05:21,478 -    base.py -    INFO: Iter:          5, Max. Viol.:   0.381688, Time: 0:00:14.259243
[Rank 0] 2017-03-30 15:05:24,262 -    base.py -    INFO: Iter:          6, Max. Viol.:   0.277613, Time: 0:00:17.042925
[Rank 0] 2017-03-30 15:05:27,043 -    base.py -    INFO: Iter:          7, Max. Viol.:   0.249915, Time: 0:00:19.824264
[Rank 0] 2017-03-30 15:05:29,809 -    base.py -    INFO: Iter:          8, Max. Viol.:   0.243094, Time: 0:00:22.590356
[Rank 0] 2017-03-30 15:05:32,575 -    base.py -    INFO: Iter:          9, Max. Viol.:   0.191051, Time: 0:00:25.356007
[Rank 0] 2017-03-30 15:05:35,337 -    base.py -    INFO: Iter:         10, Max. Viol.:   0.150705, Time: 0:00:28.117966
[Rank 0] 2017-03-30 15:06:01,239 -    base.py -    INFO: Iter:         19, Max. Viol.:   0.000000, Time: 0:00:54.020156
[Rank 0] 2017-03-30 15:06:01,239 -    base.py -    INFO: Optimization stopped after 19 iterations.
[Rank 0] 2017-03-30 15:06:01,368 -    base.py -    INFO: Sparsity of W is: 0.0594432
[Rank 0] 2017-03-30 15:06:01,368 -    base.py -    INFO: Sparsity and sum of alpha are: 0.0109703 - 4607.21
[Rank 0] 2017-03-30 15:06:01,896 -     run.py -    INFO: Training time: 54.772 s
[Rank 0] 2017-03-30 15:06:01,896 -     run.py -    INFO: Storing model.
[Rank 0] 2017-03-30 15:06:02,449 -     run.py -    INFO: Serializing time: 0.553 s



In [3]:

    
! $PYTHON ../scripts/run.py ww_sparse --test_data $TEST_DATA --model_dir ./model









    



[Rank 0] 2017-03-30 15:06:45,414 -     run.py -    INFO: Loading model.
[Rank 0] 2017-03-30 15:06:45,414 -    base.py -    INFO: Solver uses testing mode.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
[Rank 0] 2017-03-30 15:06:45,469 -    base.py -    INFO: Using 1 MPI processes and 1 MP threads.
[Rank 0] 2017-03-30 15:06:45,906 -     run.py -    INFO: Unserializing time: 0.491 s
[Rank 0] 2017-03-30 15:06:45,906 -     run.py -    INFO: Loading test data.
[Rank 0] 2017-03-30 15:06:46,226 -     run.py -    INFO: Start prediction.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
[Rank 0] 2017-03-30 15:06:46,226 -    base.py -    INFO: Using 1 MPI processes and 1 MP threads.
[Rank 0] 2017-03-30 15:06:46,852 -     run.py -    INFO: Test time: 0.626 s
[Rank 0] 2017-03-30 15:06:46,852 -     run.py -    INFO: Accuracy: 0.452099

Training and predicting in one shot

The following code line trains the model on the training data and then predicts on the test data.



In [4]:

    
! $PYTHON ../scripts/run.py ww_sparse --train_data $TRAIN_DATA --test_data $TEST_DATA --options folds=5 group_count=64 -v 1









    



[Rank 0] 2017-03-30 15:06:51,488 -    base.py -    INFO: Solver uses testing mode.
[Rank 0] 2017-03-30 15:06:51,542 -     run.py -    INFO: Loading train data.
[Rank 0] 2017-03-30 15:06:52,353 -     run.py -    INFO: Start training.
[Rank 0] 2017-03-30 15:06:52,353 -    base.py -    INFO: Use C=1 and epsilon=0.1
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
[Rank 0] 2017-03-30 15:06:52,353 -    base.py -    INFO: Using 1 MPI processes and 1 MP threads.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.ww <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.ww' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/ww.so'>
[Rank 0] 2017-03-30 15:06:52,378 -    base.py -    INFO: Remove 0 samples where K is 0.
[Rank 0] 2017-03-30 15:06:52,413 -      ww.py -    INFO: Use 64 groups.
[Rank 0] 2017-03-30 15:06:52,449 -    base.py -    INFO: Mem allocations:
[Rank 0] 2017-03-30 15:06:52,450 -    base.py -    INFO:   Total: 522.870 MB
[Rank 0] 2017-03-30 15:06:55,577 -    base.py -    INFO: Iter:          1, Max. Viol.:   2.273799, Time: 0:00:03.127347
[Rank 0] 2017-03-30 15:06:59,072 -    base.py -    INFO: Iter:          2, Max. Viol.:   2.343154, Time: 0:00:06.621902
[Rank 0] 2017-03-30 15:07:01,963 -    base.py -    INFO: Iter:          3, Max. Viol.:   1.059850, Time: 0:00:09.512923
[Rank 0] 2017-03-30 15:07:04,788 -    base.py -    INFO: Iter:          4, Max. Viol.:   0.730385, Time: 0:00:12.337820
[Rank 0] 2017-03-30 15:07:07,719 -    base.py -    INFO: Iter:          5, Max. Viol.:   0.381688, Time: 0:00:15.268811
[Rank 0] 2017-03-30 15:07:10,634 -    base.py -    INFO: Iter:          6, Max. Viol.:   0.277613, Time: 0:00:18.183486
[Rank 0] 2017-03-30 15:07:13,429 -    base.py -    INFO: Iter:          7, Max. Viol.:   0.249915, Time: 0:00:20.978478
[Rank 0] 2017-03-30 15:07:16,213 -    base.py -    INFO: Iter:          8, Max. Viol.:   0.243094, Time: 0:00:23.763388
[Rank 0] 2017-03-30 15:07:18,995 -    base.py -    INFO: Iter:          9, Max. Viol.:   0.191051, Time: 0:00:26.544906
[Rank 0] 2017-03-30 15:07:21,792 -    base.py -    INFO: Iter:         10, Max. Viol.:   0.150705, Time: 0:00:29.342025
[Rank 0] 2017-03-30 15:07:47,349 -    base.py -    INFO: Iter:         19, Max. Viol.:   0.000000, Time: 0:00:54.898683
[Rank 0] 2017-03-30 15:07:47,349 -    base.py -    INFO: Optimization stopped after 19 iterations.
[Rank 0] 2017-03-30 15:07:47,508 -    base.py -    INFO: Sparsity of W is: 0.0594432
[Rank 0] 2017-03-30 15:07:47,509 -    base.py -    INFO: Sparsity and sum of alpha are: 0.0109703 - 4607.21
[Rank 0] 2017-03-30 15:07:48,189 -     run.py -    INFO: Training time: 55.836 s
[Rank 0] 2017-03-30 15:07:48,189 -     run.py -    INFO: Loading test data.
[Rank 0] 2017-03-30 15:07:48,717 -     run.py -    INFO: Start prediction.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
[Rank 0] 2017-03-30 15:07:48,718 -    base.py -    INFO: Using 1 MPI processes and 1 MP threads.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.ww <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.ww' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/ww.so'>
[Rank 0] 2017-03-30 15:07:49,423 -     run.py -    INFO: Test time: 0.705 s
[Rank 0] 2017-03-30 15:07:49,423 -     run.py -    INFO: Accuracy: 0.452099

Training with two threads

The following command trains with two threads.

Compared to the execution with one thread above the running time reduces from 55.836 s to 34.166 s.



In [5]:

    
! $PYTHON ../scripts/run.py ww_sparse --train_data $TRAIN_DATA --test_data $TEST_DATA --options nr_threads=2 folds=5 group_count=64 -v 1









    



[Rank 0] 2017-03-30 15:08:39,709 -    base.py -    INFO: Solver uses testing mode.
[Rank 0] 2017-03-30 15:08:39,768 -     run.py -    INFO: Loading train data.
[Rank 0] 2017-03-30 15:08:40,543 -     run.py -    INFO: Start training.
[Rank 0] 2017-03-30 15:08:40,543 -    base.py -    INFO: Use C=1 and epsilon=0.1
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
[Rank 0] 2017-03-30 15:08:40,543 -    base.py -    INFO: Using 1 MPI processes and 2 MP threads.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.ww <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.ww' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/ww.so'>
[Rank 0] 2017-03-30 15:08:40,568 -    base.py -    INFO: Remove 0 samples where K is 0.
[Rank 0] 2017-03-30 15:08:40,597 -      ww.py -    INFO: Use 64 groups.
[Rank 0] 2017-03-30 15:08:40,632 -    base.py -    INFO: Mem allocations:
[Rank 0] 2017-03-30 15:08:40,632 -    base.py -    INFO:   Total: 522.870 MB
[Rank 0] 2017-03-30 15:08:42,376 -    base.py -    INFO: Iter:          1, Max. Viol.:   2.273799, Time: 0:00:01.743913
[Rank 0] 2017-03-30 15:08:44,061 -    base.py -    INFO: Iter:          2, Max. Viol.:   2.343154, Time: 0:00:03.428325
[Rank 0] 2017-03-30 15:08:45,768 -    base.py -    INFO: Iter:          3, Max. Viol.:   1.059850, Time: 0:00:05.135508
[Rank 0] 2017-03-30 15:08:47,451 -    base.py -    INFO: Iter:          4, Max. Viol.:   0.730385, Time: 0:00:06.818456
[Rank 0] 2017-03-30 15:08:49,139 -    base.py -    INFO: Iter:          5, Max. Viol.:   0.381688, Time: 0:00:08.506632
[Rank 0] 2017-03-30 15:08:50,786 -    base.py -    INFO: Iter:          6, Max. Viol.:   0.277613, Time: 0:00:10.153579
[Rank 0] 2017-03-30 15:08:52,428 -    base.py -    INFO: Iter:          7, Max. Viol.:   0.249915, Time: 0:00:11.795652
[Rank 0] 2017-03-30 15:08:54,102 -    base.py -    INFO: Iter:          8, Max. Viol.:   0.243094, Time: 0:00:13.469609
[Rank 0] 2017-03-30 15:08:55,827 -    base.py -    INFO: Iter:          9, Max. Viol.:   0.191051, Time: 0:00:15.194482
[Rank 0] 2017-03-30 15:08:57,544 -    base.py -    INFO: Iter:         10, Max. Viol.:   0.150705, Time: 0:00:16.911303
[Rank 0] 2017-03-30 15:09:14,030 -    base.py -    INFO: Iter:         19, Max. Viol.:   0.000000, Time: 0:00:33.397958
[Rank 0] 2017-03-30 15:09:14,031 -    base.py -    INFO: Optimization stopped after 19 iterations.
[Rank 0] 2017-03-30 15:09:14,170 -    base.py -    INFO: Sparsity of W is: 0.0594432
[Rank 0] 2017-03-30 15:09:14,170 -    base.py -    INFO: Sparsity and sum of alpha are: 0.0109703 - 4607.21
[Rank 0] 2017-03-30 15:09:14,708 -     run.py -    INFO: Training time: 34.166 s
[Rank 0] 2017-03-30 15:09:14,709 -     run.py -    INFO: Loading test data.
[Rank 0] 2017-03-30 15:09:15,091 -     run.py -    INFO: Start prediction.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
[Rank 0] 2017-03-30 15:09:15,091 -    base.py -    INFO: Using 1 MPI processes and 2 MP threads.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.ww <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.ww' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/ww.so'>
[Rank 0] 2017-03-30 15:09:15,747 -     run.py -    INFO: Test time: 0.656 s
[Rank 0] 2017-03-30 15:09:15,748 -     run.py -    INFO: Accuracy: 0.452099

Training with two processes

The following command trains with two processes.

Mpi needs to be installed for this example.

Compared to the execution with one thread above the running time reduces from 55.836 s to 36.284 s.



In [6]:

    
! mpiexec -n 2 $PYTHON ../scripts/run.py ww_sparse --train_data $TRAIN_DATA --test_data $TEST_DATA --options folds=5 group_count=64 -v 1









    



[Rank 0] 2017-03-30 15:09:34,291 -    base.py -    INFO: Solver uses testing mode.
[Rank 0] 2017-03-30 15:09:34,352 -     run.py -    INFO: Loading train data.
[Rank 0] 2017-03-30 15:09:35,235 -     run.py -    INFO: Start training.
[Rank 0] 2017-03-30 15:09:35,235 -    base.py -    INFO: Use C=1 and epsilon=0.1
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
[Rank 0] 2017-03-30 15:09:35,235 -    base.py -    INFO: Using 2 MPI processes and 1 MP threads.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.ww <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.ww' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/ww.so'>
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.ww <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.ww' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/ww.so'>
[Rank 0] 2017-03-30 15:09:35,364 -    base.py -    INFO: Remove 0 samples where K is 0.
[Rank 0] 2017-03-30 15:09:35,402 -      ww.py -    INFO: Use 64 groups.
[Rank 0] 2017-03-30 15:09:35,454 -    base.py -    INFO: Mem allocations:
[Rank 0] 2017-03-30 15:09:35,454 -    base.py -    INFO:   Total: 401.959 MB
[Rank 0] 2017-03-30 15:09:37,268 -    base.py -    INFO: Iter:          1, Max. Viol.:   1.860290, Time: 0:00:01.806321
[Rank 0] 2017-03-30 15:09:39,018 -    base.py -    INFO: Iter:          2, Max. Viol.:   1.865952, Time: 0:00:03.556403
[Rank 0] 2017-03-30 15:09:40,835 -    base.py -    INFO: Iter:          3, Max. Viol.:   0.812660, Time: 0:00:05.372441
[Rank 0] 2017-03-30 15:09:42,691 -    base.py -    INFO: Iter:          4, Max. Viol.:   0.449393, Time: 0:00:07.228644
[Rank 0] 2017-03-30 15:09:44,549 -    base.py -    INFO: Iter:          5, Max. Viol.:   0.326411, Time: 0:00:09.087122
[Rank 0] 2017-03-30 15:09:46,389 -    base.py -    INFO: Iter:          6, Max. Viol.:   0.202817, Time: 0:00:10.927235
[Rank 0] 2017-03-30 15:09:48,254 -    base.py -    INFO: Iter:          7, Max. Viol.:   0.266039, Time: 0:00:12.791925
[Rank 0] 2017-03-30 15:09:50,064 -    base.py -    INFO: Iter:          8, Max. Viol.:   0.164717, Time: 0:00:14.601591
[Rank 0] 2017-03-30 15:09:51,908 -    base.py -    INFO: Iter:          9, Max. Viol.:   0.158098, Time: 0:00:16.446103
[Rank 0] 2017-03-30 15:09:54,261 -    base.py -    INFO: Iter:         10, Max. Viol.:   0.146283, Time: 0:00:18.798770
[Rank 0] 2017-03-30 15:10:10,778 -    base.py -    INFO: Iter:         17, Max. Viol.:   0.000000, Time: 0:00:35.315510
[Rank 0] 2017-03-30 15:10:10,778 -    base.py -    INFO: Optimization stopped after 17 iterations.
[Rank 0] 2017-03-30 15:10:10,968 -    base.py -    INFO: Sparsity of W is: 0.0982164
[Rank 0] 2017-03-30 15:10:10,968 -    base.py -    INFO: Sparsity and sum of alpha are: 0.0109878 - 4601.15
[Rank 0] 2017-03-30 15:10:11,519 -     run.py -    INFO: Training time: 36.284 s
[Rank 0] 2017-03-30 15:10:11,519 -     run.py -    INFO: Loading test data.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.ww <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.ww' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/ww.so'>
[Rank 0] 2017-03-30 15:10:12,334 -     run.py -    INFO: Start prediction.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.base <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.base' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/base.so'>
[Rank 0] 2017-03-30 15:10:12,334 -    base.py -    INFO: Using 2 MPI processes and 1 MP threads.
<type 'numpy.float64'> <type 'numpy.uint64'> .dtype_f64_idtype_ui64.ww <module 'xcsvm.solvers.cython.dtype_f64_idtype_ui64.ww' from '/home/alber/.pyxbld_testing/lib.linux-x86_64-2.7/xcsvm/solvers/cython/dtype_f64_idtype_ui64/ww.so'>
[Rank 0] 2017-03-30 15:10:12,949 -     run.py -    INFO: Test time: 0.615 s
[Rank 0] 2017-03-30 15:10:12,949 -     run.py -    INFO: Accuracy: 0.452637